Towards realistic simulation of facial deformation in sign language

Abstract

Facial expressions are important aspects in any face to face communication, but they are considered more crucial in sign language (SL) communication since they can convey both emotional and linguistic information. In order to improve the hearing impaired people access to information and communication in their daily lives, many researchers have paid a great deal of interest on facial expressions synthesis from 3D facial models for rendering interactive signing avatars, while others works have focused on the human face gestures synthesis from images of real living persons. We present in this paper an automatic approach for the generation of sign language animation from a two dimensional frontal image. Our contribution consists essentially in creating intermediate frames from the input single image to generate fluid facial gestures and this by dividing the face model in little triangles and deforming them afterward by using an affine transformation.